Analyses of multivariately determined community matrices using cluster analysis and multidimensional scaling

Linear discriminant analysis was used to compute the p-variate MACARTHUR-LEVINS and Density Overlap measures of niche overlap between all pairs of 24 passerine bird species. The overlap values for the species pairs were then organized into a community matrix for each approach. The relationships inherent in the community matrices were structured by cluster analysis and non-metric multidimensional scaling. Cluster analysis identified the highly related species groups whereas multidimensional scaling demonstrated community wide relationships. In particular, the scaling approach clearly delineated shrub density and ground cover gradients.